How do you assist robots construct higher robots? By simulating much more robots.
NVIDIA founder and CEO Jensen Huang right this moment showcased how main electronics producer Quanta is utilizing AI-enabled robots to examine the standard of its merchandise.
In his keynote speech at this week’s COMPUTEX commerce present in Taipei, Huang introduced on how electronics producers are digitalizing their state-of-the artwork factories.
For instance, robots from Quanta subsidiary Techman Robotic tapped NVIDIA Isaac Sim — a robotics simulation utility constructed on NVIDIA Omniverse — to develop a customized digital twin utility to enhance inspection on the Taiwan-based electronics supplier’s manufacturing line.
The under demo reveals how Techman makes use of Isaac Sim to optimize the inspection of robots by robots on the manufacturing line. In impact, it’s robots constructing robots.
Automated optical inspection, or AOI, helps producers extra shortly determine defects and ship high-quality merchandise to their prospects across the globe. The NVIDIA Metropolis imaginative and prescient AI framework, now enabled for AOI, can also be used to optimize inspection workflows for merchandise starting from cars to circuit boards.
Techman developed AOI with its factory-floor robots through the use of Isaac Sim to simulate, take a look at and optimize its state-of-the-art collaborative robots, or cobots, whereas utilizing NVIDIA AI and GPUs for coaching within the cloud and inference on the robots themselves.
Isaac Sim is constructed on NVIDIA Omniverse — an open improvement platform for constructing and working industrial metaverse functions.
Distinctive options of Techman’s robotic AOI options embrace their placement of the inspection digital camera immediately on articulated robotic arms and GPUs built-in within the robotic controller.
This enables the bots to examine areas of merchandise that mounted cameras merely can’t entry, in addition to use AI on the edge to immediately detect defects.
“The distinctive options of Techman’s robots — in comparison with different robotic manufacturers — lie of their built-in imaginative and prescient system and AI inference engine,” mentioned Scott Huang, chief operations officer at Techman. “NVIDIA RTX GPUs energy up their AI efficiency.”
However programming the motion of those robots might be time consuming.
A developer has to find out the exact arm positions, in addition to probably the most environment friendly sequence, to seize doubtlessly a whole bunch of photos as shortly as potential.
This will contain a number of days of effort, exploring tens of hundreds of prospects to find out an optimum answer.
The answer: robotic simulation.
Utilizing Omniverse, Techman constructed a digital twin of the inspection robotic — in addition to the product to be inspected — in Isaac Sim.
Programming the robotic in simulation decreased time spent on the duty by over 70%, in comparison with programming manually on the true robotic. Utilizing an correct 3D mannequin of the product, the applying might be developed within the digital twin even earlier than the true product is manufactured, saving helpful time on the manufacturing line.
Then, with highly effective optimization instruments in Isaac Sim, Techman explored a large variety of program choices in parallel on NVIDIA GPUs.
The tip consequence was an environment friendly answer that decreased the cycle time of every inspection by 20%, in line with Huang.
Each second saved in inspection time will drop all the way down to the underside line of Techman’s manufacturing prospects.
Gathering and labeling real-world photos of defects is dear and time consuming, so Techman turned to artificial information to enhance the standard of inspections. It used the Omniverse Replicator framework to shortly generate high-quality artificial datasets.
These completely labeled photos are used to coach the AI fashions within the cloud and dramatically improve their efficiency.
And dozens of AI fashions might be run on the edge — effectively and with low latency due to NVIDIA expertise — whereas inspecting notably difficult merchandise, a few of which take greater than 40 fashions to scrutinize their totally different features.
Be taught extra about how Isaac Sim on Omniverse, Metropolis and AI are streamlining the optical inspection course of throughout merchandise and industries by becoming a member of NVIDIA at COMPUTEX, the place the Techman cobots shall be on show.